Solar power plants are large-scale infrastructures that require regular inspection operations, to verify that each panel is functioning as desired. Thermal imaging is often used since variations in the panel temperature can be used to recognize a possible failure. Drones can be applied, being ideally suited for monitoring large areas in a short time: a drone equipped with an infrared camera is thus flown over each panel. Visual data from a standard camera can also be integrated into the analysis. However, these operations can still be time-consuming for larger plants; moreover, a skilled operator is needed during the entire inspection, and it is not obvious how to optimize the route over the whole set of panels (also taking into account battery constraints). A semi- or fully-automated system is thus of practical interest. Here, we present our current work (in collaboration with a company that performs such inspections), proposing an automated drone system for solar plants. First, a satellite picture is automatically downloaded (knowing the coordinates of the plant); this image is analyzed by a state-of-the-art machine learning algorithm that detects the positions of the panel lines. Then, the software finds the optimal route passing over each panel, by solving a Travelling Salesman Problem. This route is followed by the drone over the plant, employing a visual servoing algorithm. We have gathered preliminary results from both simulations and tests in our laboratory: we successfully demonstrated both the route planning phase and the visual servoing algorithm.

Inspection of Large-Scale Solar Plants by an Autonomous Drone: Planning and Control

Fabio Conti;Giovanni Mottola;Carmine Recchiuto;Antonio Sgorbissa
2023-01-01

Abstract

Solar power plants are large-scale infrastructures that require regular inspection operations, to verify that each panel is functioning as desired. Thermal imaging is often used since variations in the panel temperature can be used to recognize a possible failure. Drones can be applied, being ideally suited for monitoring large areas in a short time: a drone equipped with an infrared camera is thus flown over each panel. Visual data from a standard camera can also be integrated into the analysis. However, these operations can still be time-consuming for larger plants; moreover, a skilled operator is needed during the entire inspection, and it is not obvious how to optimize the route over the whole set of panels (also taking into account battery constraints). A semi- or fully-automated system is thus of practical interest. Here, we present our current work (in collaboration with a company that performs such inspections), proposing an automated drone system for solar plants. First, a satellite picture is automatically downloaded (knowing the coordinates of the plant); this image is analyzed by a state-of-the-art machine learning algorithm that detects the positions of the panel lines. Then, the software finds the optimal route passing over each panel, by solving a Travelling Salesman Problem. This route is followed by the drone over the plant, employing a visual servoing algorithm. We have gathered preliminary results from both simulations and tests in our laboratory: we successfully demonstrated both the route planning phase and the visual servoing algorithm.
File in questo prodotto:
File Dimensione Formato  
Conti_IRIM_2023.pdf

accesso aperto

Tipologia: Documento in Pre-print
Dimensione 11.42 MB
Formato Adobe PDF
11.42 MB Adobe PDF Visualizza/Apri
Conti_IRIM_2023_compressed.pdf

accesso aperto

Tipologia: Documento in versione editoriale
Dimensione 139.27 kB
Formato Adobe PDF
139.27 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11567/1174099
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact